METHOD AND DEVICE FOR DETERMINING FRAUDULENT TRANSACTIONS OF USER Russian patent published in 2025 - IPC G06Q20/40 G06N3/45 G06N3/08 

FIELD: physics.

SUBSTANCE: disclosed solution relates to anti-fraud solutions. Method of determining fraudulent transactions of a user comprises steps of: obtaining a sequence of user transactions for a given time interval, wherein each transaction is characterized by a set of attributes; converting each received transaction into an attribute vector based on the corresponding set of transaction attributes; processing the obtained attribute vectors using a machine learning coding model (MLCM), during which a sequence of attribute vectors is obtained in the form of a matrix of hidden states; processing a matrix of hidden states using a generative machine learning model (GMLM), trained on matrices of hidden states of sequences of vectors of transaction attributes, during which the GMLM from the matrix of hidden states generates related vectors of attributes of one or more transactions; generating the last user transaction based on the generated attribute vector; comparing the last perfect transaction of the user and the generated last transaction of the user by calculating the anomaly score of the completed transaction; generating a fraudulent transaction signal if the value of the abnormality estimate is greater than the threshold value.

EFFECT: high completeness and accuracy of determining fraudulent transactions.

5 cl, 4 dwg